Current Issue Cover
基于质点的可变形体自碰撞检测

纪传舜1, 刘卉1(复旦大学计算机科学技术学院,上海 200433)

摘 要
自碰撞检测是可变形体模拟过程中最耗时的环节,提出一种使用图形硬件的快速算法。算法以质点而非三角形作为自碰撞检测的基本单元,用球体包围以质点为中心的局部区域,再用AABB包围该球体的运动轨迹并将数据组织成纹理送入GPU,通过两遍离屏渲染计算出碰撞对集合及每个碰撞对的碰撞发生时间,算法复杂度为O(n)。实验结果表明,使用该算法在大规模布料模拟中检测自碰撞,效率较高。
关键词
Particle-based self collision detection for deformable objects using graphics hardware

()

Abstract
Self collision detection is the most time consuming process in deformable object simulation. In this paper, a novel algorithm is presented for performing self collision using graphics hardware. Each particle, rather than the traditional triangle, is taken as a primitive. The particle-centered local regions are bounded with a sphere, where the AABB is then enclosed. After initial culling, the AABBs of potentially colliding particles are organized in textures and sent to GPU. A linear time two-pass rendering procedure is proposed to compute the set of colliding pairs and the colliding time of each pair. Experimental results demonstrate this algorithm handles self collision at interactive rates in the context of cloth simulation.
Keywords

订阅号|日报